Download Turning Over a Golden Leaf? Global Liquidity and Emerging Market

INDIAN INSTITUTE OF MANAGEMENT
AHMEDABAD  INDIA
Research and Publications
Turning Over a Golden Leaf?
Global Liquidity and Emerging Market Central Banks’
Demand for Gold after the Financial Crisis
Balagopal Gopalakrishnan and Sanket Mohapatra
W. P. No. 2017-04-02
April 2017
The main objective of the working paper series of the IIMA is to help faculty members, research staff and doctoral
students to speedily share their research findings with professional colleagues and test their research findings at the
pre-publication stage. IIMA is committed to maintain academic freedom. The opinion(s), view(s) and conclusion(s)
expressed in the working paper are those of the authors and not that of IIMA.
INDIAN INSTITUTE OF MANAGEMENT
AHMEDABAD-380 015
INDIA
Research and Publications
Turning Over a Golden Leaf?
Global Liquidity and Emerging Market Central Banks’ Demand for
Gold after the Financial Crisis
Balagopal Gopalakrishnan and Sanket Mohapatra
Indian Institute of Management, Ahmedabad
Abstract
The quantity of gold reserves held by central banks in emerging markets and developing
economies (EMDEs) has risen sharply following the global financial crisis in 2008. This paper
examines factors driving holding of gold by central banks in 50 EMDEs using a dynamic panel
generalized method of moments model. We find that monetary expansion in advanced
economies is robustly related to the post-crisis increase in EMDE gold reserves, after controlling
for domestic factors and changes in the global risk environment. This effect holds across
different measures of global liquidity, and is robust to alternate model specifications, inclusion of
additional covariates, and alternate estimation methods. We argue that the unprecedented
monetary expansion in advanced economies has resulted in a shift in EMDE reserve asset
holding strategy, resulting in continued accumulation of gold reserves even after the peak of the
financial crisis.
Keywords: gold reserves, emerging markets, central banks, monetary easing
JEL Codes: G11, F31, F33
_________________________
* We thank Ricardo Correa, Jamus Jerome Lim and Arvind Sahay for useful comments and suggestions. The views
expressed in this paper are solely those of the authors and do not represent the views of the Indian Institute of
Management, Ahmedabad. Contact emails: [email protected], [email protected].
W. P. No. 2017-04-02
Page No. 2
“The desire of gold is not for gold. It is for the means of freedom and benefit.”
- Ralph Waldo Emerson
1. Introduction
The gold holdings of central banks in emerging markets and developing countries (EMDEs1)
mostly followed a similar trajectory as advanced economies for several decades until the mid2000s, as gold was gradually supplemented or replaced by a variety of hard currencies such as
the U.S. dollar, British pounds, yen, and euro (O’Connor et al, 2015). With the onset of the
global financial crisis (GFC) or the Great Recession in 2008, there was a sharp increase in the
gold holdings of central banks. This portfolio shift towards gold during the global crisis reflected
its status as a safe asset at a time of intense financial turmoil, as has been documented for other
“risk-off” periods and crises episodes (Baur & Lucey, 2010; Baur & McDermott, 2010; WGC
2016). Gold as a safe haven asset is attributed to its role as one of the first forms of money that
was used traditionally as an inflation hedge, and lack of correlation with other types of assets, an
important differentiator in a globalized world where the most of the asset classes are highly
correlated (Baur & Lucey, 2010). However, as pointed out by Aizenman & Inoue (2013), in an
era when ‘plastic money’ and ‘electronic money’ are increasingly accepted as the norm for
providing intermediation services, the role of precious commodity metals such as gold in central
bank reserves remains a puzzle.
During the post-crisis years, gold holdings of advanced economies fell in tandem with a decline
in global risk. However, gold holdings of EMDE central banks have continued to rise in this
1
EMDE classification is based on the World Bank country classification. We have considered UMIC, LMIC and
LIC income group classification as EMDE countries for the study.
W. P. No. 2017-04-02
Page No. 3
period (figure 1). This sharp increase in the post-crisis period after several decades of relative
stability suggests a structural change in EMDE central bank gold holdings. Overall gold reserves
of EMDE central banks rose from 129.8 million troy ounces in 2005 to 167 million ounces by
the end of 2009, and further to 242.1 million ounces by 2015 – an 86 percent increase from the
level in 2005. Statistical analysis using a Markov regime switching test (Hamilton, 1989)
indicates a structural break in aggregate EMDE gold holdings during 2009-2015 compared to the
previous period (see appendix figure A.1). This aggregate pattern is broadly mirrored at the
country level, with some variation across individual countries.
Figure 1: Gold holdings of central banks in emerging markets and developing economies
(EMDEs) and advanced economies
Gold holdings (millions troy ounces)
1,200
250
1,000
200
800
150
600
100
400
Emerging & Developing (left axis)
Emerg. & Developing excl. China & India (left axis)
50
200
Advanced (right axis)
0
0
1980
1985
1990
1995
2000
2005
2010
2015
Source: Authors’ calculations using IMF IFS database.
The post-financial crisis period was characterized by unprecedented monetary easing in the
advanced economies, as central banks in the major G-4 countries (United States, Euro Area,
Japan, and the United Kingdom) resorted to a range of conventional and unconventional
W. P. No. 2017-04-02
Page No. 4
monetary policies to limit the effect of the GFC on output and employment (Fawley & Neely,
2013). The unconventionality of the monetary policies implemented by these central banks can
be evidenced using a regime switching test, which shows that central bank assets experienced a
post-crisis structural break (see Appendix figure A.2). The results indicate a regime change in the
way monetary measures have been used to enhance liquidity in the system.
The heads of several EMDE central banks had expressed apprehensions about the spillover
effects of these measures by G-4 central banks on their currencies through massive capital
inflows, and potential vulnerability to sudden reversals of these flows (Eichengreen, 2013;
Rajan, 2015). Some EMDE policymakers deemed it a “currency war” 2 in which advanced
economies were trying to undervalue the respective currencies to boost their economic activity,
which led to a lack of trust amongst the G-20 nations (Zoellick, 2011). In addition, these
measures also acted as a signal to the EMDE central banks on the outlook of the G4 currency
denominated assets.
The massive and persistent monetary easing for several years since the GFC likely led to a shift
in the reserve accumulation strategy of EMDE central banks away from the so-called hard
currencies. As shown in figure 2, the share of gold in EMDE central bank reserves has risen
together with a corresponding fall in the share of the G-4 major currencies.3 Figure 2 provides
two estimates of the proportion of G-4 currency reserves to overall reserves (allocated reserves
from the COFER database) including gold reserves of the EMDE countries reporting to IMF.
2
Guido Mantega, finance minister of Brazil in 2010 publicly asserted that the world was in the midst of an
international currency war. https://www.ft.com/content/33ff9624-ca48-11df-a860-00144feab49a
3
We also observe a sharp increase in the share of other currencies held by EMDE central banks, from about 2
percent of overall reserves in 2008 to 8 percent in 2015. This supports our main hypothesis that EMDE central have
diversified away from G-4 currencies to gold and other currencies. IMF COFER does not provide disaggregated data
on the holding of currencies at the country level; hence, we choose to focus on gold where such country-level data is
available from the World Gold Council.
W. P. No. 2017-04-02
Page No. 5
The first estimate (Alt. 1 in figure 2) is based on the gold reserves of those EMDE countries that
report their reserve composition to IMF, and the other estimate (Alt. 2) is based on the
cumulative gold reserves of all the EMDE central banks. The trends indicate a dramatic shift in
the proportion of reserves held as gold and in G-4 currencies after 2008, indicating a significant
shift in the way the reserve composition of EMDE central banks has evolved.
Figure 2: Composition of overall EMDE central bank reserves 1999-2015
98.0%
9.0%
96.0%
8.0%
94.0%
7.0%
92.0%
6.0%
90.0%
5.0%
88.0%
4.0%
86.0%
3.0%
G4 currencies % of total reserves, Alt. 1 (left axis)
84.0%
G4 currencies % of total reserves, Alt. 2 (left axis)
2.0%
82.0%
Gold % total reserves, Alt. 1 (right axis)
1.0%
Gold % of total reserves, Alt. 2 (right axis)
2015
2014
2013
2012
2011
2010
2009
2008
2007
2006
2005
2004
* total reserves includes gold
2003
2002
2001
2000
0.0%
1999
80.0%
Source: Authors’ calculations using IMF COFER database and IMF IFS.
Note: Alt. 1 is a conservative estimate of G-4 currency holdings in overall reserves of EMDEs, and Alt. 2 is a liberal
estimate (see text). Currencies other the G-4 currencies account for the remaining part of overall reserves.
In this paper, we empirically investigate whether and how excessive monetary easing in
advanced economies, which is largely exogenous to EMDEs, may have influenced the regime
shift in gold accumulation by EMDE central banks. In order to estimate this, we use a dynamic
panel generalized method of moments model to understand whether excessive monetary easing
by the central banks in the advanced economies can explain this increase in EMDE gold
reserves, after controlling for domestic factors and changes in the global risk environment. We
use several different proxies for studying the impact of this massive loosening in liquidity
W. P. No. 2017-04-02
Page No. 6
conditions on the change in gold reserves composition of EMDE central banks. The results
indicate that the monetary easing in advanced economies which followed the GFC played a
significant role in the demand for gold reserves by EMDE central banks. We validate our
baseline results by conducting robustness checks using different specifications, inclusion of
additional covariates and alternate estimation techniques.
This work contributes to the literature in several ways. To our knowledge, this is the first study
to analyze the shift in gold accumulation strategy of EMDE central bank after the GFC. Our
findings reiterate the role of gold as a safe haven, particularly for EMDE countries due to the
perceived policy uncertainty in the post-crisis years. This research complements the work done
by Aizenman & Inoue (2013) for advanced economy central banks by looking at EMDE central
banks exclusively.
The next section discusses the theoretical background behind shifts in reserve accumulation
strategies of EMDE central banks. This is followed by data and methodology in Section 3 and a
discussion of results and analysis in Section 4. The subsequent section concludes with some
implications for policy.
2. Theoretical background: Role of liquidity in shifts in reserve accumulation
strategies of central banks
Coordination problems among economies originate from the asymmetries prevalent in the
structures, growth rate and relative cyclical positions of the developed and developing economies
(Obstfeld 2013). Greater the cooperation among the competing economies from both developed
and developing side, lesser the need for international liquidity to mitigate the imbalances.
W. P. No. 2017-04-02
Page No. 7
According to Obstfeld, a lack of cooperation among the nations has caused excess liquidity to be
injected into the IMS post the GFC of 2008, thereby revealing the extent to the imbalances
prevalent today.
The dominance of the US dollar in the international monetary system (IMS) as the main reserve
currency following the Bretton woods era has perpetuated the so-called Triffin dilemma (Triffin,
1960). The Triffin dilemma in its earlier form was in the context of the Bretton woods agreement
of convertibility of US dollar to gold. Obstfeld (2013) draws attention to the reserve holdings
status of the central banks of all countries in 1970, which stood at USD 42 billion as against the
gold holding of United States, which was USD 11 billion, indicating that the US had to cover the
shortfall in claims by buying up more gold should there have been a run at that point in time. The
shortfall was such that even if the treasury could purchase all of the official monetary gold and
the supply of gold, it would have fallen short; this would have resulted in a potential fiscal
insolvency of the US.
The modern day Triffin dilemma discussed by Farhi, Gourinchas, & Rey (2011) argues that the
governments in the core of the IMS will have to continue the issuance of gross government debt
to satisfy the demand for reserves from the countries in the periphery. The present day dilemma
remains even in a multipolar world with several reserve currencies (Obstfeld, 2013). A
supplementary problem with this dilemma is that the increasing external debt of the countries’
issuing reserve currencies will undermine the confidence of the investors in the reserve asset
(Lee, 2014).
Miranda-Agrippino & Rey (2015a) argue that US monetary policy is the primary driver of the
global financial cycles and questions the level of independence enjoyed by other central banks,
W. P. No. 2017-04-02
Page No. 8
even those with flexible exchange rates such as the UK or the ECB.4 Therefore, the massive
monetary easing conducted by the advanced economy central banks, led by the US Federal
Reserve, have increased the vulnerabilities of the EMDE central banks given the spillover effects
on the EMDE countries, evidenced in several studies (Fratzscher, Lo Duca, & Straub, 2016; Lim
& Mohapatra, 2016).
Another reason for the shift in reserve accumulation strategy of emerging market economies is
the internationalization of some of the currencies that are increasingly considered as reserve
currencies and a call for a multipolar IMS. As mentioned in recent literature, China has shown its
intent in making its currency fully convertible (Chen & Cheung, 2011; Fratzscher & Mehl,
2014). In the same vein, there is evidence that while the dominance of US dollar continues postGFC, the pillars supporting that role is waning (Kirshner, 2014). However, there are challenges
to full convertibility of yuan since the country is attempting to orchestrate a transition without
hampering domestic growth and export growth (Chan, 2017; Lo, 2016). The delay in the
convertibility of yuan increases the uncertainty surrounding its potential reserve currency status
as well.
The implications of the massive expansion of G-4 central bank balance sheets for future financial
stability and risks stemming from a revelation of underlying imbalances can potentially cause a
loss of confidence in the existing major reserve currencies. The need for diversification by the
EMDE economies stem from these imbalances and the challenges faced by China in making
yuan fully convertible. Gold is perceived to be a safe asset during uncertain and highly volatile
times, and this characteristic of gold is evidenced in the literature (O’Connor, Lucey, Batten, &
4
Miranda-Agrippino & Rey (2015a) find that one global factor in the form of US monetary policy explains an
important part of the variance in asset returns across the world.
W. P. No. 2017-04-02
Page No. 9
Baur, 2015). The role of gold in portfolio diversification has been found in many studies
(Beckmann, Berger, & Czudaj, 2015; Bredin, Conlon, & Poti, 2015) and policy makers have
called for active management of gold in central bank portfolio to help in diversification
(European Central Bank, 2004). Given the benefits offered by gold, EMDE central banks have
resorted to an accumulation of gold reserves to hedge the potential risks that are prevalent in the
IMS since the GFC.
3. Data and Methodology
3.1 Data description
An annual database of EMDE central banks’ gold reserves and the other global as well as
country-specific variables was constructed. The EMDE gold reserves data was from the IMF
International Financial Statistics (IFS) database. All the other variables were obtained from the
World Development Indicators (WDI), IMF International Financial Statistics, Chicago Board of
Options Exchange (CBOE) database, Federal Reserve Bank of St. Louis database (FRED), and
World Bank International debt statistics (see Appendix table A.1 for variable description and
sources). Appendix table A.2 presents the list of 50 EMDE countries in our sample, and
appendix table A.3 presents pairwise correlations among the variables.
Liquidity indicators can be broadly classified as price based and quantity based indicators
(Adrian & Shin, 2009). Among quantity based indicators, the conventional measure of liquidity
is the monetary aggregate of the advanced economies (Belke, Bordon, & Volz, 2013; Chung,
Lee, Loukoianova, Park, & Shin, 2015) which is considered the core liquidity measure. The
other quantity indicator that has gained importance post the global financial crisis is the credit
W. P. No. 2017-04-02
Page No. 10
based aggregates that is accounted by the official liquidity, which is created by the central banks
through both conventional and unconventional measures, and the private liquidity created by
financial institutions (Committee on the Global Financial System, 2011). Conventional price
based indicator of liquidity is the short term interest rate of the advanced economies (Adrian &
Shin, 2009).
As a measure of this liquidity, the variables used for the analysis include a Global Liquidity
Measure (GLM), G-4 central bank asset holding (G4CB)5, GDP-weighted short-term rate (STR)
of the advanced economies, and US treasury purchases (USLM). GLM is taken in constant
dollars as well as in real terms as indicated below:
∑ (
)
∑ (
(1)
)
(2)
where i refers to the country under study in time period t and MA refers to monetary aggregate in
local currency units, LCU_USD_2010 is the currency conversion to 2010 US dollars and P is the
CPI in the respective year. GLMC and GLMR refer to the liquidity measure in US dollars at
constant 2010 exchange rates and in real (inflation-adjusted) local currency converted to US
dollars at constant exchange rate. These variables account for fluctuations in monetary
aggregates that may arise from exchange rate fluctuations among the major currencies, as well as
from domestic inflation. A similar method is used to arrive at the aggregate central banking
5
The G-4 advanced economies comprise the Euro Area (19 countries), Japan, United States, and the United
Kingdom
W. P. No. 2017-04-02
Page No. 11
assets of the G-4 advanced economies (G4CBC and G4CBR) in constant USD and real USD
terms respectively. Descriptive statistics for these liquidity measures are presented in table 1.
Global risk has been recognized as a determinant of the gold holdings of central banks in a study
done by Gopalakrishnan & Mohapatra (2017). Global risk is also considered to be driving the
global financial cycle (Miranda-Agrippino & Rey, 2015b). VIX index, a measure of risk used in
the literature to indicate global uncertainty, is used in this study to capture global risk component
(Garcia, Ortiz, & Cowan, 2006; Jubinski, Lipton, & Joseph, 2013).
Table 1: Descriptive statistics
Variable
Global
Log of GLMC
Log of GLMR
Log of G4CBC
Log of G4CBR
GDP-weighted short term rate (%)
TED Spread (%)
Log of US treasury purchases real
VIX index
USD appreciation
Country-specific
Gold reserves as share of GDP (%)
GDP growth (%)
Trade openness to GDP (%)
Capital Account openness
Private credit to GDP (%)
Log of GDP per capita
Log of IIR
CPI Inflation (%)
Log of Gold price (Const. 2010 USD)
External debt to GDP (%)
Inflation volatility
Mean
Std. Dev.
Min
Max
31.0163
30.9949
29.2468
29.2739
2.29889
0.49706
20.9301
21.4135
0.0018
0.2606
0.2934
0.4918
0.4341
1.6696
0.3514
0.8069
6.1247
0.0477
30.5757
30.5131
28.5863
28.7200
0.2501
0.1900
19.998
12.800
-0.0596
31.3928
31.4338
30.1475
30.0828
5.7628
1.5500
22.2493
32.690
0.1258
0.97601
4.38768
76.8406
41.1635
37.1183
7.55858
3.51577
9.08952
6.51601
50.3781
0.00925
1.2564
4.0404
35.292
32.036
31.023
1.0642
0.4780
16.627
0.5384
33.824
0.0214
0.0000
-14.759
15.4195
0.0000
1.2669
4.6636
1.6864
-72.729
5.8462
4.0807
0.0000
8.9693
55.556
220.230
100.000
160.1250
9.6514
4.3845
293.733
7.34709
260.482
0.3998
The country-specific determinants of gold reserves include capital account openness, trade
openness, GDP per capita, GDP growth, external debt to GDP, the ratio of private credit to
GDP, gold price (in 2010 US dollars), Institutional Investor Risk (IIR) index, inflation and
W. P. No. 2017-04-02
Page No. 12
inflation volatility. Higher capital account openness usually increases the vulnerability of the
country to sudden changes in gross capital flows (Roveda & Rosselli, 2003). Both capital
account openness and trade openness are measures used as determinants of central bank reserve
holdings in previous studies (Lane & Burke, 2001; Marc-Andre & Parent, 2005; Obstfeld,
Shambaugh, & Taylor, 2009). GDP per capita is a proxy for economic size, which is one of the
determinants of the overall reserves. GDP growth factors in the propensity to accumulate gold
reserves and also controls for the strategic initiatives by many countries to ensure consistent
growth (Aizenman & Lee, 2008; Benigno & Fornaro, 2012; Cheng, 2015).
Private credit to GDP, an indicator for financial development (Svirydzenka, 2016), has been
used as one of the determinants of reserve holdings by central banks (Lane & Burke, 2001).
Institutional Investors Rating (IIR) is used as a proxy for country-specific risk perceptions.
Inflation and foreign currency debt are considered to affect the overall gold holdings of a central
bank (Aizenman & Inoue, 2013). Gold price in real terms is used to control for the effect on the
value of gold reserves, however, we also evaluate the baseline regressions without gold price in
the robustness section. The last control variable added to the baseline specification is the US
dollar appreciation (computed from the US dollar index against a basket of its trading partners).
It is well documented that the commodity prices are inversely related to the US dollar (Akram,
2009; Hammoudeh, Nguyen, & Sousa, 2015).
3.2. Methodology
The specification for the determinants of gold holdings by emerging market central banks can be
expressed as:
W. P. No. 2017-04-02
Page No. 13
(3)
where G is gold reserves as a share of GDP in EMDE countries, GLM is the global liquidity
measure which is based on monetary policy measures in advanced economies, Z is a vector of
global factors other than GLM, X is a vector of macroeconomic and institutional variables at the
country-level,
is a country-specific intercept, and
is an i.i.d error term.
As discussed in the earlier section, the vector of global variables includes VIX, international gold
price (in 2010 US dollars), and US dollar appreciation. The country-specific variables include
capital account openness, trade openness, GDP per capita, GDP growth, IIR country ratings, gold
price, inflation, private credit to GDP, foreign currency external debt to GDP, and inflation
volatility. In some alternate specifications, we bring in TED spread, net portfolio flow volatility
and long term yields as additional covariates. Summary statistics are shown in table 2.
It has been evidenced that the coefficients obtained from ordinary least squares with fixed effects
are prone to dynamic panel bias (Nickell, 1981) in cases where one of the explanatory variables
has a lag of the dependent variable. We control for this bias using a dynamic panel generalized
method of moments (GMM) estimators. The estimator used in this study is the Systems GMM
(SGMM) estimator of Arellano & Bover (1995), which is suitable for datasets with lower time
periods and a larger number of entities under study. SGMM uses the first difference and level
residuals of the estimate for the moment conditions and is considered efficient given the
flexibility in the use of instruments (Roodman, 2009).
Another econometric issue that we foresee in the specification is the potential endogeneity of
some of the explanatory variables to central banks’ gold holding and these issues are
W. P. No. 2017-04-02
Page No. 14
acknowledged in related literature such as Bussière, Cheng, Chinn, & Lisack (2015) for GDP
growth, Baur (2016) for gold price, capital account openness, and IIR. This issue is mitigated by
using the SGMM estimator which permits and accounts for potential endogenous variables.
Moreover, we use one period lagged explanatory variable to avoid contemporaneous feedback
effects.
4. Results
4.1 Baseline results
We analyze the effect of liquidity measures undertaken by advanced economy central banks
using two global liquidity measures (GLM) as indicated in the earlier section: GLMC and
GLMR6. This measure of liquidity, based on monetary aggregates, is the conventional measure
of liquidity used in a lot of studies. The results obtained from the dynamic panel regression
method are as shown in table 2 below. In order to understand the impact of the unconventional
monetary measures, the estimations were done using the overall sample period (1999 to 2015)
and two sub-periods (1999-2007 and 2008 to 2015) to identify the effect of the structural change
in the responsiveness to the unconventional monetary measures.
The overall results using the GLMC and GLMR measures are as shown in columns (1) and (4) of
table 2, and the sensitivity of the gold holdings of EMDE central banks are positive and
significant to the GLM measures after controlling for other factors, both domestic and global,
that determine the accumulation of gold reserves. This effect is larger since 2008 (columns (3)
6
A bivariate analysis of the median gold holdings (gold_to_gdp) and the GLMR is as shown in appendix
figures A.3.
W. P. No. 2017-04-02
Page No. 15
and (6)) indicating a shift in the accumulation strategy from then on and coincides with the
liquidity loosening undertaken by the advanced economies.
The results of the period until 2007 in columns (2) and (5) of table 2 indicate that there was
hardly any significant relationship between the liquidity measures and gold accumulation of the
EMDE central banks. Sensitivity of gold accumulation to global liquidity for the entire time
period, which is positive and significant, is driven by the changes in gold accumulation post
2008 with no significant effect from pre-crisis period.
In addition to the GLM measures, we evaluate the role of the central bank assets in the G4
countries as a proxy for the excess liquidity and its effect on the gold accumulation of the EMDE
central banks. The results are as shown below in table 3. The sensitivities of gold holdings to this
measure both in constant USD (G4CBC) and real USD terms (G4CBR) is consistent with the
earlier results for GLM.
As evidenced in table 3, the effect of liquidity is not significant pre-2008 and the overall results are
driven by the post crisis developments for the period starting 2008 (columns (3) and (6)). The post
2008 sensitivity of the gold reserves is higher for GLMR measure of liquidity than the G4CBR
measure. The results obtained in this analysis substantiate the role of monetary easing, conducted
by G4 central banks through multiple channels, in determining the gold holdings of EMDE central
banks.
W. P. No. 2017-04-02
Page No. 16
Table 2: Global liquidity and central bank gold reserves in EMDEs
Global liquidity measure – const. USD
(GLMC)
Gold to GDPt-1
Log GLMCt-1
Full
sample
(1)
Pre-2008
sample
(2)
Post-2008
sample
(3)
Full
sample
(4)
Pre-2008
sample
(5)
Post-2008
sample
(6)
0.983***
(0.116)
1.341**
(0.528)
0.990***
(0.099)
0.362
(0.362)
0.874***
(0.157)
2.811***
(0.787)
0.851***
(0.085)
0.936***
(0.085)
0.946***
(0.142)
0.483
(0.584)
-0.005
(0.004)
1.431***
(0.503)
0.023
(0.020)
-0.005*
(0.003)
0.005*
(0.003)
0.004
(0.004)
0.021
(0.126)
-0.23
(0.244)
0.000
(0.005)
-0.116
(0.168)
-0.001
(0.002)
7.446
(8.328)
-13.523
(16.772)
5.245***
(1.312)
0.027***
(0.007)
-4.202***
(1.021)
-0.026*
(0.014)
-0.004
(0.004)
-0.001
(0.004)
-0.005
(0.004)
0.044
(0.298)
0.044
(0.481)
-0.012
(0.013)
-0.833***
(0.301)
0.004
(0.005)
-5.159***
(1.767)
-158.255***
(40.330)
0.014***
(0.003)
-0.916**
(0.372)
-0.01
(0.012)
-0.005
(0.004)
0.001
(0.002)
-0.001
(0.003)
-0.18
(0.133)
0.284
(0.263)
-0.001
(0.004)
-0.518**
(0.238)
0.000
(0.001)
-6.746
(6.263)
-37.693***
(14.577)
-0.006
(0.004)
1.394***
(0.465)
0.019
(0.018)
-0.004**
(0.002)
0.005**
(0.002)
0.004
(0.003)
0.027
(0.099)
-0.257
(0.223)
-0.001
(0.004)
-0.158
(0.158)
-0.001
(0.002)
7.033
(10.278)
-9.449
(10.043)
0.030***
(0.009)
-4.324***
(1.051)
-0.032**
(0.015)
-0.004
(0.003)
-0.001
(0.005)
-0.004
(0.005)
-0.119
(0.280)
0.169
(0.532)
-0.02
(0.016)
-0.650*
(0.349)
0.004
(0.006)
-5.896
(6.118)
-82.713***
(23.685)
1.382**
(0.631)
0.011***
(0.003)
-0.740**
(0.369)
-0.011
(0.012)
-0.004
(0.004)
0.001
(0.002)
0.000
(0.003)
-0.068
(0.120)
0.166
(0.202)
0.000
(0.004)
-0.342*
(0.184)
0.000
(0.001)
-5.239
(4.741)
-40.563**
(18.059)
Yes
667 (50)
Yes
335 (50)
Yes
332 (50)
Yes
667 (50)
Yes
335 (50)
Yes
332 (50)
52
43.159
-0.553
656.489
39
26.788
-0.513
673.48
44
36.435
-1.519
155.909
55
42.744
-0.648
842.482
40
27.446
-0.556
668.615
44
33.402
-1.19
173.47
Log GLMRt-1
VIX t-1
USD appreciationt-1
GDP growtht-1
Trade opennesst-1
KAOPENt-1
Private credit to GDPt-1
Log GDP per capitat-1
Log IIRt-1
Inflationt-1
Log Gold pricet-1
Debt to GDP t-1
Inflation Vol t-1
Constant
Country fixed-effects
No. of obs. (countries)
No. of instruments
Hansen statistic
AR(2) statistic
Wald Chi-square
Global liquidity measure – Real
(GLMR)
Notes: The dependent variable is central bank gold reserves as a percentage of GDP for EMDE countries. GLMC is the global
liquidity measure at constant 2010 exchange rates. GLMR is the global liquidity measure in real (inflation-adjusted) terms and at
constant 2010 exchange rates. Standard errors presented below the coefficient estimates in parenthesis are Windmeijer corrected and
clustered at the country level. *** Significant at 1%, ** significant at 5% and * significant at 10%. All the SGMM regressions pass
the Hansen test of over-identifying restrictions and the test for second-order autocorrelation suggesting that the SGMM moment
conditions are not violated.
W. P. No. 2017-04-02
Page No. 17
Table 3: Central bank assets of G-4 advanced economies and gold reserves in
EMDEs
G-4 Central bank assets - const. USD
(G4CBC)
Gold to GDPt-1
Log G4CBCt-1
All
years
(1)
Pre-2008
sample
(2)
Post-2008
sample
(3)
All
years
(4)
Pre-2008
sample
(5)
Post-2008
sample
(6)
0.840***
(0.079)
0.285*
(0.154)
0.949***
(0.088)
-0.027
(0.193)
0.927***
(0.157)
1.030***
(0.240)
0.819***
(0.070)
0.949***
(0.088)
0.922***
(0.163)
-0.066
(0.221)
-0.005
(0.003)
1.435***
(0.508)
0.055
(0.132)
1.170***
(0.269)
0.032***
(0.007)
-4.410***
(0.937)
-1.031***
(0.339)
Yes
Yes
335 (50)
40
26.379
-0.652
649.342
Yes
Yes
332 (50)
44
35.727
-1.057
170.153
0.014***
(0.003)
-0.792*
(0.408)
-0.068
(0.120)
-0.006
(0.004)
1.425***
(0.509)
0.037
(0.133)
0.033***
(0.007)
-4.307***
(0.945)
-1.002***
(0.326)
0.329**
(0.151)
0.014***
(0.003)
-0.802*
(0.430)
-0.069
(0.112)
Yes
Yes
667 (50)
57
39.869
-0.855
862.877
Yes
Yes
335 (50)
40
26.299
-0.63
633.729
Yes
Yes
332 (50)
44
35.654
-1.067
171.576
Yes
Yes
667 (50)
53
37.195
-0.84
1027.036
Log G4CBRt-1
VIX t-1
USD appreciationt-1
Log Gold pricet-1
Country controls
Country fixed-effects
No. of obs. (countries)
No. of instruments
Hansen statistic
AR(2) statistic
Wald Chi-square
G-4 Central bank assets - Real
(G4CBR)
Notes: The dependent variable is central bank gold reserves as a percentage of GDP for EMDE countries. G4CBC is the combined
central bank assets in the G-4 advanced economies (Euro Area, Japan, United States, and the United Kingdom) at constant 2010
exchange rates. G4CBR is the combined central assets of G-4 economies in real (inflation-adjusted) terms and at constant 2010
exchange rates. Standard errors presented below the coefficient estimates in parenthesis are Windmeijer corrected and clustered at
the country level. *** Significant at 1%, ** significant at 5% and * significant at 10%. All the SGMM regressions pass the Hansen
test of over-identifying restrictions and the test for second-order autocorrelation suggesting that the SGMM moment conditions are
not violated.
4.2 Alternate measures of global liquidity
In this section, we estimate the above baseline specification using alternate measures of liquidity
other than the GLM and G4CB variables. The variables used in the analysis are a US liquidity
measure, which is the cumulative treasury purchases (in real 2010 US dollars) by the US Federal
Reserve since 2002, and the GDP-weighted short-term interest rate of the advanced economies.
The use of alternative measures helps us in verifying our findings for the global measure and
W. P. No. 2017-04-02
Page No. 18
whether the measures undertaken by the leading advanced economies such as the United States
are in line with the overall results.
The results of this analysis are as shown in columns (1), (2) and (3) of table 4. The finding here
is that there is an increase in the sensitivity of gold holdings of EMDE central banks since 2008
when the US Federal Reserve decided to start the large-scale asset purchase (LSAP) programs to
improve the liquidity in the system. The findings for the U.S liquidity measure are similar to the
global liquidity measure with the sensitivity being positive and significant post crisis and in the
same order of magnitude as the cumulative G4CBR.
Table 4: Alternative measures of global liquidity
US Liquidity Measure (USLM)
Gold to GDPt-1
Log USLMt-1
All
years
(1)
Pre-2008
sample
(2)
Post-2008 sample
(3)
0.811***
(0.086)
0.048
(0.049)
0.908***
(0.096)
-0.165
(0.118)
0.900***
(0.153)
0.457**
(0.221)
All
years
(4)
Pre-2008
sample
(5)
Post-2008
sample
(6)
0.039***
(0.012)
-3.453***
(0.853)
-0.974**
(0.409)
0.814***
(0.071)
-0.011
(0.011)
-0.021*
(0.013)
0.011***
(0.004)
-0.503
(0.361)
-0.047
(0.106)
0.952***
(0.060)
-0.003
(0.011)
0.002
(0.012)
-0.010**
(0.004)
0.767**
(0.329)
-0.029
(0.121)
0.887***
(0.124)
-0.194***
(0.048)
-0.114**
(0.044)
0.011**
(0.005)
-3.435***
(0.829)
-0.922**
(0.364)
0.013***
(0.004)
-0.811
(0.513)
0.134
(0.084)
-0.004
(0.004)
0.92
(0.657)
0.156
(0.130)
Yes
Yes
545 (50)
50
38.537
-0.763
520.011
Yes
Yes
213 (50)
36
17
-0.476
1160.855
Yes
Yes
332 (50)
44
28.746
-0.466
266.811
Yes
Yes
706 (50)
51
41.123
-1.119
1225.343
Yes
Yes
374 (50)
44
33.615
-0.648
2195.493
Yes
Yes
332 (50)
46
38.013
-1.197
316.349
STRt-1
VIX t-1
USD appreciationt-1
Log Gold pricet-1
Country controls
Country fixed-effects
No. of obs. (countries)
No. of instruments
Hansen statistic
AR(2) statistic
Wald Chi-square
Short-term interest rates (STR)
Notes: The dependent variable is central bank gold reserves as a percentage of GDP for EMDE countries. USLM is the liquidity
measure for the United States in real (inflation-adjusted) terms and at constant 2010 exchange rates. STR is the weighted-average
of short-term interest rates in the G-4 advanced economies. Standard errors presented below the coefficient estimates in
parenthesis are Windmeijer corrected and clustered at the country level.*** Significant at 1%, ** significant at 5% and *
significant at 10%. All the SGMM regressions pass the Hansen test of over-identifying restrictions and the test for second-order
autocorrelation suggesting that the SGMM moment conditions are not violated.
W. P. No. 2017-04-02
Page No. 19
Next, we estimate the baseline results using a GDP-weighted short-term rate measure for the
countries that were used to arrive at GLM. This is a price indicator of liquidity in the system and
the advanced economy central banks have consistently reduced the base rates since the crisis
going to zero lower bounds and in the case of some countries, going beyond the zero lower
bounds to negative interest rate territories.
As indicated in columns (4), (5) and (6) in table 4, the results indicate that the gold holdings in
EMDE central banks have risen in response to lowering short-term rates across the advanced
countries. This is consistent with the results we obtained in terms of monetary aggregates and
central bank asset holdings.
4.3 Robustness tests
In the initial set of robustness checks, we test the consistency of the results after adding
additional covariates. One of the possibilities is that the sudden capital flows and stops due to the
advanced economy central bank decisions would have made the portfolio flows into the EMDE
countries more volatile. We control for this by taking the volatility of portfolio flows for the
preceding three years as an explanatory variable. The results in table 5, columns (1)-(3), are
consistent with the baseline results.
Next, we test whether the addition of TED spread as a proxy for global risk in addition to VIX
(columns (4)-(6) of table 5) or if the addition of long-term US yield, a proxy for long-term rates
in the advanced countries (columns (7)-(9)), affects our baseline estimations. The results
indicating a stronger relationship between global liquidity and EMDE central bank gold holdings
in the post-crisis period are consistent with our baseline model.
W. P. No. 2017-04-02
Page No. 20
Table 5: Global liquidity measure (GLMR) with additional covariates
Gold to GDPt-1
Log GLMRt-1
VIXt-1
USD appreciationt-1
Log NPFVt-1
With Net
Portfolio flow
volatility
(Overall)
(1)
With Net
Portfolio flow
volatility (Pre2008)
(2)
With Net
Portfolio flow
volatility
(Post-2008)
(3)
With
TED
Spread
(overall)
(4)
With
TED
Spread
(Pre-2008)
(5)
0.880***
(0.110)
1.724***
(0.659)
0.015***
(0.004)
-0.723*
(0.410)
0.012
(0.023)
0.935***
(0.046)
0.65
(0.460)
-0.005
(0.003)
1.332***
(0.473)
-0.025
(0.021)
0.957***
(0.101)
4.762***
(1.457)
0.032***
(0.008)
-3.699***
(1.020)
0.044
(0.061)
0.876***
(0.096)
1.359**
(0.678)
0.013***
(0.003)
-0.63
(0.387)
0.954***
(0.114)
-0.318
(1.030)
-0.006
(0.005)
1.475*
(0.827)
0.905***
(0.127)
5.078***
(1.436)
0.032***
(0.008)
-3.912***
(1.084)
0.025
(0.040)
-0.077
(0.211)
-0.183
(0.188)
TED spreadt-1
With
TED
Spread
(Post-2008)
(6)
US 10 year treasury yieldt-1
With
US 10 year
yield (Overall)
(7)
With
US 10 year
yield
(Pre-2008)
(8)
0.929***
(0.088)
1.845***
(0.629)
0.011***
(0.003)
-0.905**
(0.457)
0.982***
(0.094)
0.719
(1.549)
-0.004
(0.006)
1.13
(0.826)
0.975***
(0.156)
5.462**
(2.527)
0.031***
(0.008)
-3.863*
(2.320)
0.127***
(0.039)
-0.195
(0.173)
0.016
(0.087)
-0.198
(0.400)
0.085
(0.318)
-0.805
(0.695)
With
US 10 year yield
(Post-2008)
(9)
Log Gold pricet-1
-0.367*
(0.199)
-0.244
(0.188)
-0.905***
(0.350)
-0.361*
(0.214)
0.088
(0.286)
-0.841**
(0.336)
Country controls
Country fixed-effects
No. of obs. (countries)
No. of instruments
Hansen statistic
AR(2) statistic
Wald Chi-square
Yes
Yes
590(46)
Yes
Yes
287(46)
Yes
Yes
303(46)
Yes
Yes
667 (50)
Yes
Yes
335(50)
Yes
Yes
332(50)
Yes
Yes
667 (50)
Yes
Yes
335(50)
Yes
Yes
332(50)
49
35.651
-0.083
458.128
37
23.358
-0.532
1282.393
44
34.42
-1.553
260.581
57
42.315
-0.692
1182.855
45
28.954
-0.727
488.157
47
40.011
-1.424
224.396
53
38.027
-0.783
807.591
41
25.765
-0.558
531.475
45
30.583
-0.796
292.888
Notes: The dependent variable is central bank gold reserves as a percentage of GDP for EMDE countries. GLMR is the global liquidity measure in real (inflation-adjusted) terms and at constant 2010
exchange rates. Standard errors presented below the coefficient estimates in parenthesis are Windmeijer corrected and clustered at the country level. *** Significant at 1%, ** significant at 5% and *
significant at 10%. All the SGMM regressions pass the Hansen test of over-identifying restrictions and the test for second-order autocorrelation suggesting that the SGMM moment conditions are not
violated.
In addition, we check the robustness of the results using the following subsample and alternate
specifications. We exclude the five BRICS (Brazil, Russia, India, China and South Africa)
nations from the analysis and see if the results are consistent for the rest of the developing
countries. The results as shown in in table 6 columns (1), (2) and (3) indicate that the results are
consistent with the baseline results with full sample. Although the coefficient for post-crisis
period is lower than the full sample indicating a relatively lower sensitivity of non-BRICS
countries to global liquidity measures, the results suggest that central banks in other developing
nations have also shifted towards the accumulation of gold post-crisis.
In the next set of robustness checks, we estimate the results by removing gold price from the
specification (as shown in columns (4) to (6) of table 6) and treat VIX as an endogenous variable
(as shown in columns (7) to (9)). Although central banks account for a relatively small fraction
of overall global gold demand, it is possible that changes in central bank gold reserve holdings
may influence international gold prices at the margin. The direction of the sensitivity and the
significance levels of the GLM measure for the post-crisis period are consistent with the baseline
results.
The final robustness test that we do is to construct an alternative GLMR measure including the
BRICS countries’ monetary aggregates. We treat this measure as endogenous in the specification
and analyze the sensitivity of this measure. The results (see columns (10) to (12)) indicate that
the sensitivity for the post-crisis period is statistically significant and is in the hypothesized
direction.
Table 6: Global liquidity measure (GLMR) with subsamples and alternate specifications
Gold to GDPt-1
Log GLMRt-1
VIX t-1
USD appreciationt-1
Log Gold pricet-1
Country controls
Country fixedeffects
No. of obs.
(countries)
No. of instruments
Hansen statistic
AR(2) statistic
Wald Chi-square
Without
Gold
Price All
years
(4)
0.866***
(0.070)
0.620**
(0.241)
0.012***
(0.003)
-0.352
(0.364)
Without
Gold
Price Pre2008
(5)
0.932***
(0.083)
0.103
(0.377)
-0.005*
(0.003)
1.461***
(0.429)
Without
Gold
Price
Post-2008
(6)
0.751***
(0.132)
2.787**
(1.321)
0.022***
(0.007)
-3.229***
(0.888)
With
VIX as
endog.
All years
(7)
0.994***
(0.102)
1.449**
(0.596)
0.009***
(0.003)
-0.417
(0.327)
-0.362*
(0.201)
With
VIX as
endog.
Pre-2008
(8)
0.995***
(0.060)
0.443
(0.382)
-0.006**
(0.003)
1.426***
(0.430)
-0.078
(0.122)
With
VIX as
endog.
Post-2008
(9)
0.898***
(0.132)
4.545***
(1.246)
0.030***
(0.007)
-4.089***
(0.997)
-0.938***
(0.353)
GLM with
BRICS
monetary
aggregates
All year
(10)
0.899***
(0.093)
1.173**
(0.473)
0.017***
(0.004)
-0.935**
(0.406)
-0.499**
(0.198)
GLM with
BRICS
monetary
aggregates
Pre-2008
(11)
0.943***
(0.060)
0.089
(0.360)
-0.005*
(0.003)
1.532***
(0.451)
-0.131
(0.210)
GLM with
BRICS
monetary
aggregates
Post-2008
(12)
0.900***
(0.134)
2.553***
(0.753)
0.037***
(0.009)
-3.557***
(0.993)
-0.945**
(0.424)
Excl.
BRICS
All years
(1)
0.808***
(0.098)
1.444**
(0.640)
0.014***
(0.004)
-0.705*
(0.395)
-0.342*
(0.176)
Excl.
BRICS
Pre-2008
(2)
0.973***
(0.062)
0.348
(0.468)
-0.006
(0.004)
1.377***
(0.502)
-0.128
(0.186)
Excl.
BRICS
Post-2008
(3)
0.885***
(0.105)
3.944***
(0.964)
0.031***
(0.008)
-3.508***
(0.804)
-0.692**
(0.271)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
592 (50)
52
39.338
-0.53
712.683
295 (50)
39
30.012
-0.594
582.003
297 (50)
42
35.354
-1.433
286.177
667 (50)
50
42.124
-0.843
767.532
335(50)
38
26.777
-0.513
548.78
332(50)
43
27.225
-1.935
267.963
667 (50)
54
41.093
-0.596
544.739
335 (50)
38
26.154
-0.563
813.039
332 (50)
45
35.681
-1.02
238.205
667 (50)
51
43.319
-0.329
443.483
335 (50)
42
30.467
-0.527
1343.283
332 (50)
47
38.674
-1.437
323.456
Notes: The dependent variable is central bank gold reserves as a percentage of GDP for EMDE countries. GLMR is the global liquidity measure in real (inflation-adjusted) terms and at constant 2010
exchange rates. Standard errors presented below the coefficient estimates in parenthesis are Windmeijer corrected and clustered at the country level. *** Significant at 1%, ** significant at 5% and *
significant at 10%. All the SGMM regressions pass the Hansen test of over-identifying restrictions and the test for second-order autocorrelation suggesting that the SGMM moment conditions are not
violated.
Finally we check the robustness of the results with alternative estimation techniques. The results
obtained using the Driscoll & Kraay (1998) technique, where the standard errors are
heteroskedasticity-consistent as well as control for cross-sectional dependence, is in the
hypothesized direction for the post crisis period (see table 7, columns (1) to (3)). The last test in
this section uses a bias corrected estimation technique in which the initialization is done using an
Anderson & Hsiao (1982) estimator. The results obtained support the post crisis dependence of
gold reserves on the global liquidity measure (see columns (4) to (6)).
Table 7: Global liquidity measure (GLMR) with alternate estimation techniques
Driscoll
Kraay
All years
(1)
Driscoll
Kraay
Pre-2008
(2)
Driscoll
Kraay
Post-2008
(3)
2.749**
(1.299)
0.022***
(0.003)
-3.593***
(0.934)
0.707***
(0.120)
Bias corrected
LSDV
All years
(4)
0.675***
(0.031)
0.644*
(0.344)
0.014***
(0.002)
-0.614
(0.383)
-0.074
(0.112)
Bias corrected
LSDV
Pre-2008
(5)
0.714***
(0.053)
0.023
(0.421)
-0.001
(0.005)
1.241
(0.501)
0.105
(0.218)
Bias corrected
LSDV
Post-2008
(6)
0.522***
(0.051)
3.710***
(1.121)
0.026***
(0.004)
-3.285***
(0.794)
-0.045
(0.182)
-0.256
(0.516)
0.010*
(0.005)
-0.392
(1.069)
0.722***
(0.179)
0.095
(0.208)
-0.003
(0.005)
1.319***
(0.128)
0.353**
(0.168)
Yes
Yes
667 (50)
Yes
Yes
335(50)
Yes
Yes
332(50)
Yes
Yes
667 (50)
Yes
Yes
335 (50)
Yes
Yes
332 (50)
Gold to GDPt-1
Log GLMRt-1
VIX t-1
USD appreciationt-1
Log Gold pricet-1
Country controls
Country fixed-effects
No. of obs. (countries)
Notes: The dependent variable is central bank gold reserves as a percentage of GDP for EMDE countries. GLMR is the global
liquidity measure in real (inflation-adjusted) terms and at constant 2010 exchange rates. Standard errors presented below the
coefficient estimates in parenthesis are heteroskedasticity consistent and for Driscoll Kraay it is controlled for cross correlation
dependence as well. Bias corrected least square dummy variable estimation is initialized using an Anderson Hsiao estimator. ***
Significant at 1%, ** significant at 5% and * significant at 10%.
5. Conclusion
This study has examined the correlates of gold reserve accumulation by central banks of
emerging markets and developing economies in the years before and after the GFC. After several
years of relative stability, EMDE central bank gold reserves rose sharply during the financial
crisis and continued to increase well after the peak of the crisis. We find that the increase in
EMDE gold holdings in the post-crisis period was strongly associated with the large expansion in
liquidity and increase in central bank balance sheets in the advanced economies. This post-crisis
effect holds even after controlling for a range of country-specific factors, international gold
prices and global risk. The baseline results are robust to alternative specifications, inclusion of
additional controls and alternative estimation techniques.
We view our findings as contributing to the debates on the design of the global financial
architecture in the era of global imbalances, quantitative easing, and a massive expansion in the
balance sheets of the major central banks in the advanced economies in the years following the
GFC. Even though the US dollar continues to maintain its status as the primary global reserve
currency in the post-crisis period, the lack of credible alternatives after nearly two decades
following the introduction of the euro in 1999 and the uncertainties associated with the efforts at
internationalization of the Chinese yuan leave EMDE central bank portfolio managers with a
Hobson’s choice. The sharp increase in the demand for gold – and a parallel shift away from the
major G-4 currencies – by EMDE central banks in the post-crisis period should be seen in the
broader context of the possible risks to financial stability and the potential for revelation of
underlying imbalances that may result from the extraordinarily loose monetary policies
implemented in the advanced economies for several years after the crisis.
W. P. No. 2017-04-02
Page No. 25
References
Adrian, T., & Shin, H. S. (2009). Prices and Quantities in the Monetary Policy Transmission
Mechanism. International Journal of Central Banking.
Aizenman, J., & Inoue, K. (2013). Central banks and gold puzzles. Journal of the Japanese and
International Economies, 28, 69–90. http://doi.org/10.1016/j.jjie.2013.02.001
Aizenman, J., & Lee, J. (2008). Financial versus Monetary Mercantilism: Long‐run View of
Large International Reserves Hoarding. The World Economy, 31(5), 593–611.
Akram, Q. F. (2009). Commodity prices, interest rates and the dollar. Energy Economics, 31(6),
838–851.
Anderson, T. W., & Hsiao, C. (1982). Formulation and estimation of dynamic models using
panel data. Journal of Econometrics, 18(1), 47–82.
Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of errorcomponents models. Journal of Econometrics, 68(1), 29–51. http://doi.org/10.1016/03044076(94)01642-D
Baur, D. G. (2016). Central Banks and Gold. Available at SSRN:2326606.
http://doi.org/http://dx.doi.org/10.2139/ssrn.2326606
Baur, D. G., & Lucey, B. M. (2010). Is Gold a Hedge or a Safe Haven? An Analysis of Stocks,
Bonds and Gold. Financial Review, 45(2), 217–229. http://doi.org/10.1111/j.15406288.2010.00244.x
Beckmann, J., Berger, T., & Czudaj, R. (2015). Does gold act as a hedge or a safe haven for
stocks? A smooth transition approach. Economic Modelling, 48, 16–24.
http://doi.org/10.1016/j.econmod.2014.10.044
Belke, A., Bordon, I. G., & Volz, U. (2013). Effects of Global Liquidity on Commodity and
Food Prices. World Development, 44(July 2012), 31–43.
http://doi.org/10.1016/j.worlddev.2012.12.009
Benigno, G., & Fornaro, L. (2012). Reserve Accumulation, Growth and Financial Crises. CEP
Discussion Paper, (1161), 1–36. Retrieved from
W. P. No. 2017-04-02
Page No. 26
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2210193
Bredin, D., Conlon, T., & Poti, V. (2015). Does gold glitter in the long-run? Gold as a hedge and
safe haven across time and investment horizon. International Review of Financial Analysis,
41, 320–328. http://doi.org/10.1016/j.irfa.2015.01.010
Bussière, M., Cheng, G., Chinn, M. D., & Lisack, N. (2015). For a few dollars more: Reserves
and growth in times of crises. Journal of International Money and Finance, 52, 127–145.
http://doi.org/10.1016/j.jimonfin.2014.11.016
Chan, S. (2017). Assessing China’s recent capital outflows: policy challenges and implications.
China Finance and Economic Review, 5(1), 3.
Chen, X., & Cheung, Y. (2011). Renminbi going global. China & World Economy, 19(2), 1–18.
Cheng, G. (2015). A Growth Perspective of Foreign Reserve Accumulation. Macroeconomic
Dynamics, 19(6), 1358–1379. http://doi.org/10.1017/S1365100513000862
Chinn, M. D., & Ito, H. (2006). What matters for financial development? Capital controls,
institutions, and interactions. Journal of Development Economics, 81(1), 163–192.
http://doi.org/10.1016/j.jdeveco.2005.05.010
Chung, K., Lee, J.-E., Loukoianova, E., Park, H., & Shin, H. S. (2015). Global liquidity through
the lens of monetary aggregates. Economic Policy, 30(82), 231–290.
Committee on the Global Financial System (CGFS). (2011). Global liquidity – Concept ,
Measurement and Policy Implications. CGFS Papers 45.
Dempster, N., & Artigas, J. C. (2009). Gold as a tactical inflation hedge and Long term Strategic
Asset. World Gold Council Research Paper, (333).
Driscoll, J. C., & Kraay, A. C. (1998). Consistent covariance matrix estimation with spatially
dependent panel data. Review of Economics and Statistics, 80(4), 549–560.
Eichengreen, B. (2013). Currency war or international policy coordination? University of
California, Berkeley.
European Central Bank. (2004). Risk Management for Central Bank Foreign Reserves. (C.
Bernadell, P. Cardon, J. Coche, F. Diebold, & S. Manganelli, Eds.). European Central Bank.
W. P. No. 2017-04-02
Page No. 27
Farhi, E., Gourinchas, P.-O., & Rey, H. (2011). Reforming the international monetary system.
CEPR.
Fawley, B. W., & Neely, C. J. (2013). Four Stories of Quantitive Easing. Federal Reserve Bank
of St. Louis Review, 95(1), 51–88. Retrieved from
http://research.stlouisfed.org/publications/review/13/01/Fawley.pdf\npapers2://publication/u
uid/BC525E0D-1C8B-4C8C-8791-6E1904F96AE1
Fratzscher, M., Lo Duca, M., & Straub, R. (2016). On the international spillovers of US
quantitative easing. The Economic Journal.
Fratzscher, M., & Mehl, A. (2014). China’s Dominance Hypothesis and the Emergence of a Tri‐
polar Global Currency System. The Economic Journal, 124(581), 1343–1370.
Garcia, H. A., Ortiz, A., & Cowan, K. (2006). The Role of Global Risk Aversion in Explaining
Sovereign Spreads. Economia, 7(1), 125–155. Retrieved from
http://www.jstor.org/stable/20065508
Gopalakrishnan, B., & Mohapatra, S. (2017). Global Risk and Demand for Gold by Central
Banks (No. 2017-01-01). Indian Institute of Management Ahmedabad, Research and
Publication Department.
Hamilton, B. Y. J. D. (1989). A New Approach to the Economic Analysis of Nonstationary Time
Series and the Business Cycle. Econometrica, 57(2), 357–384.
Hamilton, J. D. (1989). A New Approach to the Economic Analysis of Nonstationary Time
Series and the Business Cycle. Econometrica, 57(2), 357–384.
http://doi.org/10.3982/ECTA5989
Hammoudeh, S., Nguyen, D. K., & Sousa, R. M. (2015). US monetary policy and sectoral
commodity prices. Journal of International Money and Finance, 57, 61–85.
Jubinski, D., Lipton, A. F., & Joseph, S. (2013). VIX , Gold , Silver , and Oil : How do
Commodities React to Financial Market Volatility ? Journal of Accounting and Finance,
13(1), 70–89.
Kirshner, J. (2014). Same as it ever was? Continuity and change in the international monetary
W. P. No. 2017-04-02
Page No. 28
system. Review of International Political Economy, 21(5), 1007–1016.
http://doi.org/10.1080/09692290.2014.946435
Lane, P. R., & Burke, D. (2001). The Empirics of Foreign Reserves. Open Economies Review,
12(4), 423–434. http://doi.org/10.1023/A:1017939118781
Lee, J. (2014). Will the Renminbi Emerge as an International Reserve Currency? The World
Economy, 37(1), 42–62.
Lim, J. J., & Mohapatra, S. (2016). Quantitative easing and the post-crisis surge in financial
flows to developing countries. Journal of International Money and Finance, 68, 331–357.
http://doi.org/10.1016/j.jimonfin.2016.02.009
Lo, C. (2016). China’s Impossible Trinity: The Structural Challenges to the “Chinese Dream.”
Springer.
Marc-Andre, G., & Parent, N. (2005). An Empirical Analysis of Foreign Exchange Reserves in
Emerging Asia (No. Bank of Canada Working Paper 2005-38). Working paper 2005-38.
Miranda-Agrippino, S., & Rey, H. (2015a). World asset markets and the global financial cycle.
National Bureau of Economic Research.
Miranda-Agrippino, S., & Rey, H. (2015b). World Asset Markets and the Global Financial
Cycle.
Nickell, S. (1981). Biases in Dynamic Models with Fixed Effects. Econometrica, 49(6), 1417–
1426. http://doi.org/10.2307/1911408
O’Connor, F. a., Lucey, B. M., Batten, J. a., & Baur, D. G. (2015). The financial economics of
gold — A survey. International Review of Financial Analysis, 41, 186–205.
http://doi.org/10.1016/j.irfa.2015.07.005
Obstfeld, M. (2013). The international monetary system: living with asymmetry. In
Globalization in an Age of Crisis: Multilateral Economic Cooperation in the Twenty-First
Century (pp. 301–336). University of Chicago Press.
Obstfeld, M., Shambaugh, J. C., & Taylor, A. M. (2009). Financial Instability, Reserves, and
Central Bank Swap Lines in the Panic of 2008. NBER Working Paper 14826. Retrieved
W. P. No. 2017-04-02
Page No. 29
from http://www.nber.org/papers/w14826
Rajan, R. (2015). Competitive Monetary Easing: Is it yesterday once more? Macroeconomics
and Finance in Emerging Market Economies, 8(1-2), 5–16.
Roodman, D. (2009). How to do xtabond2: An introduction to difference and system GMM in
Stata. Stata Journal, 9(1), 86–136. http://doi.org/The Stata Journal
Roveda, C., & Rosselli, F. (2003). The Ratio of International Reserves to Short Term External
Debt as an Indicator of External Vulnerability: Some Lessons from the Experience of
Mexico and Other Eemerging Economies.
Sari, R., Soytas, U., & Hacihasanoglu, E. (2011). Do global risk perceptions influence world oil
prices? Energy Economics, 33(3), 515–524. http://doi.org/10.1016/j.eneco.2010.12.006
Svirydzenka, K. (2016). Introducing a New Broad-based Index of Financial Development. IMF
Working Paper, (5).
Triffin, R. (1960). Gold and the dollar crisis. New Haven: Yale Univen.
Zoellick, R. (2011). The G20 Must Look Beyond Bretton Woods II. (R. Schatz & L. Watson,
Eds.)Trust Meltdown II.
W. P. No. 2017-04-02
Page No. 30
Appendix Table A.1: Data definitions and sources
Variable
Gold reserves to GDP
Global Liquidity
Measure (GLMC)
Global Liquidity
Measure (GLMR)
G4 central bank assets
(G4CBC)
G4 central bank assets
(G4CBR)
US Liquidity Measure
(USLM)
Advanced Economy
Short term rate (STR)
VIX
USD appreciation
TED spread
GDP per capita
GDP growth
Institutional Investor
Rating
Developing & High
Income countries
Private credit to GDP
Capital account
openness
Trade openness
Inflation
Gold Price
External debt to GDP
Inflation volatility
Definition and Construction
Ratio of central bank gold reserves to GDP (both values
in current US dollars)
This is a measure computed using the M2 supply for US
and Japan and M3 supply for the rest of the advanced
economies, in constant 2010 exchange rates (see text).
GLMC in constant inflation-adjusted real terms (see
text).
This is a measure computed using the G4 central bank
assets, in constant 2010 exchange rates (see text).
G4CBC in constant inflation-adjusted real terms (see
text).
Continuous measure of the US FED purchase of
treasury assets, mortgaged backed securities and federal
agency debt
GDP-weighted benchmark policy rates of all the
advanced economies.
Index of implied volatility of S&P 500 options - a
measure of global risk perception
US dollar appreciation based on the dollar index, which
is weighted average of the foreign exchange value of the
dollar against a broad group of US trading partners
The spread between interbank loans and the US
government treasury bills.
Gross domestic product per capita in nominal US dollars
Annual growth of Gross Domestic Product of a country
Institutional investor perception of the sovereign default
risk
Country classification based on income.
Domestic credit to private sector as a share of GDP
Measure of the openness of the capital account to
foreign fund flows
Ratio of total exports and imports to GDP. A measure of
the openness of the country to trade.
Measure of Consumer Price Index changes in a country
Gold price in dollars per troy ounce (2010 US dollars)
Ratio of the foreign currency debt holdings of the
country to GDP
Measure of annual volatility in CPI inflation
Data sources
IMF International Financial
Statistics (IFS)
FRED, IMF IFS, OECD
currency rates
FRED, IMF IFS, OECD
currency rates
FRED, IMF IFS, OECD
currency rates
FRED, IMF IFS, OECD
currency rates
FRED
IMF IFS
Chicago Board Options
Exchange (CBOE)
FRED
FRED
World Development
Indicators(WDI)
IMF IFS
Institutional Investor rating
index
World Bank classification
WDI
Chinn and Ito (Chinn & Ito,
2006) index
IMF Balance of Payments
statistics
IMF IFS
World Bank commodity price
data
World Bank International
Debt Statistics
IMF IFS
Appendix Table A.2: Emerging markets and developing economies (EMDE)
Albania
Bangladesh
Bolivia
Brazil
Bulgaria
Burundi
Cambodia
Cameroon
China, P.R.: Mainland
Colombia
Congo, Republic of
Costa Rica
Ecuador
Egypt
El Salvador
Fiji
Guatemala
Honduras
India
Indonesia
Jordan
Kazakhstan
Kenya
Kyrgyz Republic
Lao People's Democratic Republic
Macedonia, FYR
Malawi
Malaysia
Mauritius
Mexico
Mongolia
Morocco
Mozambique
Nepal
Nicaragua
Nigeria
Pakistan
Paraguay
Peru
Philippines
Romania
Russian Federation
South Africa
Sri Lanka
Thailand
Tunisia
Turkey
Ukraine
Vietnam
Yemen, Republic of
W. P. No. 2017-04-02
Page No. 32
Appendix table A.3: Pairwise correlations
Gold to
GDP
GLMR
GLMC
G4CBC
G4CBR
US LM
Real
STR
VIX
TED
spread
GDP
growth
Trade
openness
KAOPEN
Pvt
credit
to GDP
Log
GDPPC
Log IIR
Infla
Log
Price
Debt to
GDP
Infl
Vol
Gold to
GDP
GLMR
0.1147
1
GLMC
0.1066
0.9962
1
G4CBC
0.1179
0.9454
0.9542
1
G4CBR
US LM
Real
STR
0.1191
0.9374
0.9448
0.9995
1
0.1335
0.8551
0.8373
0.9387
0.9442
1
-0.097
-0.7515
-0.7737
-0.8159
-0.8167
-0.7395
1
VIX
-0.0099
-0.2031
-0.2461
-0.2779
-0.2742
-0.2755
-0.0741
1
TED spread
GDP
growth
Trade
openness
KAOPEN
Pvt credit to
GDP
Log
GDPPC
Log IIR
-0.0578
-0.1031
-0.156
-0.3262
-0.3409
-0.5322
0.386
0.4556
1
-0.0065
-0.0366
0.0119
-0.0521
-0.0608
-0.17
0.1097
-0.2255
0.0289
1
0.1478
0.0425
0.0629
0.0386
0.0351
-0.0209
-0.0093
-0.0544
0.0326
0.109
1
0.1152
0.0161
0.0234
0.0004
-0.0018
-0.0466
0.0006
0.0105
0.0401
-0.0657
0.0699
1
-0.0068
0.2275
0.2282
0.2307
0.23
0.1927
-0.1769
-0.0436
-0.0327
-0.013
0.292
0.0121
1
0.1389
0.4098
0.4225
0.3971
0.3927
0.2894
-0.3058
-0.0962
-0.0295
-0.0782
0.087
0.172
0.3594
0.1099
0.2108
0.2214
0.1862
0.1822
0.1074
-0.1251
-0.0476
0.0319
0.0308
0.0181
0.2314
0.5815
0.6491
1
Inflation
0.0941
-0.0797
-0.1165
-0.1039
-0.1017
-0.0872
0.1034
0.0842
0.1247
-0.0717
0.0069
-0.1291
-0.1915
-0.0917
-0.2001
Log Price
Debt to
GDP
Inflation
Vol
USD
appreciation
0.1266
0.9663
0.9594
0.9138
0.9062
0.8006
-0.7526
-0.1706
-0.1585
0.0052
0.0617
0.02
0.2179
0.422
0.2258
-0.1048
0.1053
-0.2261
-0.2526
-0.1949
-0.1884
-0.0555
0.1352
0.0286
-0.0606
-0.091
0.2242
0.0545
-0.1915
-0.3208
-0.4058
0.0707
-0.2449
1
-0.0119
0.0437
0.0325
0.0488
0.0493
0.0609
-0.0201
-0.0283
0.0042
-0.0431
0.0229
-0.0352
-0.1332
-0.1242
-0.1221
0.1007
0.0118
0.1487
1
0.0412
0.2134
0.2302
0.4051
0.4216
0.627
-0.242
0.0327
-0.1861
-0.285
-0.0566
-0.0437
0.0924
0.059
-0.0081
-0.0063
0.1193
0.0711
0.0407
USD
appre
1
1
1
1
1
Appendix figure A.1: EMDE Gold holdings since 2000
The right axis indicates the EMDE central bank gold holdings in million troy ounces. Since the time
series of gold holdings of EMDE central banks is auto correlated, an AR1 model (based on ACF plots)
and a regime switching model with unobserved state variable is fitted to the time series data of gold
holdings as per Hamilton (1989). Number of regimes chosen for the model is 2. A switch is seen in the
year 2008. The bottom graph indicates the smoothed probabilities of the model once the unobserved state
variable is estimated.
Appendix figure A.2: G4 central bank assets (trillion 2010 USD) since 2000
The right axis indicates the G4 central bank assets since 2000. Since the time series is auto correlated, an
AR1 model (based on ACF plots) and a regime switching model with unobserved state variable is fitted
to the time series data as per Hamilton (1989). Number of regimes chosen for the model is 2. A switch is
seen in the year 2008 and 2012. The bottom graph indicates the smoothed probabilities of the model once
the unobserved state variable is estimated.
W. P. No. 2017-04-02
Page No. 35